SSM - Population Health 24 (2023) 101551

Contents lists available at ScienceDirect

SSM - Population Health

journal homepage: www.elsevier.com/locate/ssmph

Partnership status and positive DNA methylation age acceleration across

the adult lifespan in the UK
Wen Wang a, Anna Dearman a, Yanchun Bao b, Meena Kumari a, *
a
Institute for Social and Economic Research, University of Essex, Wivenhoe Park, Colchester, Essex, CO4 3SQ, UK
b
Department of Mathematics, University of Essex, Wivenhoe Park, Colchester, Essex, UK

A R T I C L E I N F O A B S T R A C T

Keywords: Although a significant body of research has shown that married people are healthier and live longer, empirical
Partnership status research on sex differences in the link between marital status and health suggests results are mixed. Moreover,
DNA methylation the sex disparities in marital status and health relationships vary across adulthood. The literature on partnership
Phenoage
status and measures of ageing is largely focused on older age groups and is limited in its view of early adulthood.
DunedinPACE
Sex
Data from waves 2 and 3 (2010–2012) of Understanding Society: UKHLS were used to examine the association of
Age current partnership status with epigenetic age acceleration (AA) assessed with DNA methylation (DNAm) al­
United Kingdom gorithms ’Phenoage’ and ’ DunedinPACE ’ in 3492 participants (aged 16–97). Regression models were estimated
separately for men and women, and further stratified by age groups.
Divorced/separated and widowed people showed positive age acceleration compared to the married/cohab­
iting people (reference group). Some sex differences were apparent, especially, among the single and divorced/
separated groups. Age differences were also apparent, for example in men, being single was negatively associated
with DNAmAA in the youngest group, but positively in the oldest group compared to partnered counterparts.
These findings illustrate the importance of partnerships on the ageing process, in particular marital change
through divorce and widowhood for positive age acceleration in adults. For single groups, observations were
heterogenous by age and sex.

1. Introduction
A significant body of research has shown that married people are
healthier (both mentally and physically) and live longer compared to
their unmarried (single, divorced, or widowed) counterparts (Bulanda
et al., 2016; Carr & Springer, 2010; Hughes & Waite, 2009; Liu et al.,
2020; McFarland et al., 2013; Tatangelo et al., 2017). Marriage or being
in a partnership may be protective through the accumulation and
sharing of economic, behavioural, and psychosocial resources between
partners. However, health behaviours and socioeconomic position may
influence the association between partnership status and health. Being
in a partnership is beneficial for men and women, independent of a
variety of confounders (Brenn & Ytterstad, 2016) with some studies
suggesting a greater benefit in women than men (Wang et al., 2020).
Associations of partnership status and health can be disease specific
(Wang et al., 2020) and with broader health statuses such as frailty
suggesting associations that are not system specific (Kojima et al., 2020).
Marital disruption such as divorce may have negative impacts associated
with stress and serve to undermine health (Crowley, 2019; Liu & Waite,
2014; McFarland et al., 2013). Further, the death of a spouse is associ­
ated with a short term increased risk of mortality (Ennis & Majid, 2021;
Moon et al., 2011). As with partnership status, these associations may be
moderated by factors such as sex and age, with different associations in
different age groups (McFarland et al., 2013; Wang et al., 2020; Yu,
2023).
Explanations for the link between partnership status and health and
wellbeing have centred on the marital resource model and the stress
model by most researchers.
The marital resource model proposes an increase in health-
enhancing resources such as economic, psychosocial, and social sup­
port associated with partnership (Waite & Gallagher, 2001). For
instance, marriage or cohabitation leads to an increase in income and
wealth due to specialization, economies of scale, and the pooling of
resources within the household. Moreover, marriage/cohabitation is
associated with an increase in psychosocial resources compared to
divorce and widowhood, such as receiving support from partners and

* Corresponding author.
E-mail address: mkumari@essex.ac.uk (M. Kumari).

https://doi.org/10.1016/j.ssmph.2023.101551
Received 13 May 2023; Received in revised form 24 October 2023; Accepted 29 October 2023
Available online 5 November 2023
2352-8273/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
W. Wang et al.
stronger ties with relatives, neighbours, and shared friends (Kalmijn,
2003, 2012). The accumulation of resource advantages enhances the
opportunities to access resources that promote health, such as nutrition,
high quality living environment, healthcare, and enable healthy be­
haviours to be maintained or unhealthy behaviours to be corrected.
The stress model puts more emphasis on relationship disruption,
positing that divorce and widowhood may cause deleterious effects of
short-term stress and sustained chronic strains, which undermine peo­
ple’s health (Umberson & Thomeer, 2020; Williams & Umberson, 2004).
On the one hand, experiencing marital disruption may lead to grief, loss
of support, changes in housing and daily life, and a decline in
socio-economic position (SEP). Disadvantaged SEP reduces opportu­
nities to access resources while increasing exposure to harmful stressors
(Schreier & Chen, 2013). Response to stress may play a role in deteri­
orating health through various direct or indirect mechanisms. For
example, response to stress can alter DNA methylation and cortisol
sensitivity (Palma-Gudiel et al., 2015; Schiele et al., 2020), and sus­
tained chronic strains may result in persistent epigenetic changes (Zhang
& Liu, 2022), which may affect the ageing process and increase disease
risks. Experiencing stressful events may make individuals more likely to
adopt unhealthy behaviours (e.g., smoking and drinking) as a way of
coping, and heavy drinking and smoking are directly linked to cellular
ageing (Martins de Carvalho et al., 2019; Topiwala et al., 2022).
A number of studies suggest that the impact of relationship disso­
lution on a number of measures of health is different in men and women
(Leopold, 2018; Percheski & Meyer, 2018; Zhang et al., 2021). There are
observed patterns in sex differences in social network composition, such
that men typically receive more emotional support and favourable
regulation of health behaviours from family than women do, thus the
impact of relationship disruption on men’s health is greater than on
women (Carr & Springer, 2010; McKenzie et al., 2018; Williams &
Umberson, 2004). However, the sex disparities in marital status and
health are moderated by age. Since the normative period for some social
roles (e.g., marriage) occurs earlier for women than for men (Average
age at marriage England and Wales, by gender | Statista), the psycho­
logical and physiological disadvantage from the accumulation of marital
stress may be greater for women than men. In addition, societal changes
in patterns of partnership over time leads to different potential associ­
ations of partnership with health which is different for different age
groups. For example, as the average age of first marriages rose from 27.4
to 24.7 in 1972 to 39.7 and 37.3 by 2019 for men and women respec­
tively (Average age at marriage England and Wales, by gender |
Statista).
Previous research suggests that markers that reflect multi­
physiological systems vary by partnership status, dissolution and
widowhood (Kiecolt-Glaser, 2018; Kojima et al., 2020; Rote, 2017). At
the molecular level, evidence suggests an association of partnership
status and biomarkers of ageing such as telomere length and attrition
(Chen et al., 2020; Whisman et al., 2016; Yu & Liu, 2021). Thus, there is
potential to explore alternate molecular measures of ageing. Novel DNA
methylation algorithms of ageing have been developed as a new method
to encapsulate age related change experienced by individuals.
Ageing is an inevitable process experienced by all organisms,
occurring at molecular, cellular, organ, and organismal levels. Over the
course of biological ageing, there is compromised functionality and a
decline in the reparative and regenerative capacities of tissues and or­
gans. Such progressive loss of physiological integrity serves as the risk
factor for major human pathologies (López-Otín et al., 2013). Great ef­
forts have been undertaken to classify the molecular hallmarks of
ageing, which have a vital role in both the ageing process and
age-related diseases. In addition to telomere length, biological ageing
and rapid ageing (expressed as “age acceleration” in this study), often
calculated as a greater biological than chronological age, can be calcu­
lated using alternate methods, for example measurements have been
developed using DNA methylation (DNAm) (Bell et al., 2019), which is
the most widely studied epigenetic modification in population health.
2
SSM - Population Health 24 (2023) 101551
DNAm focuses on exploring cytosine methylation in CpG dinucleotides
(CpG methylation). Initially, DNAm age algorithms were built by iden­
tifying DNAm patterns associated with chronological age. Recently,
DNAm age algorithms, for example, the ‘Phenoage’ (Levine et al., 2018),
‘DunedinPoAm’ (Belsky et al., 2020), and ‘DunedinPACE’ (Belsky et al.,
2022) algorithms have been described. For example, the ‘Phenoage’
DNAm algorithm (Levine et al., 2018) used clinical biomarkers and
chronological age from the third National Health and Nutrition Exami­
nation Survey (NHANES) to predict a composite biological age score
(so-called "phenotypic age"). Then they used elastic net regression and
identified 513 CpG sites to predict the previously created phenotypic
age. ‘DunedinPACE’ DNAm algorithm (Belsky et al., 2020) tracked
within-individual decline over two decades in 19 multi organ system
biological indicators as the predicted phenotype, then also used elastic
net regression to identify CpG sites. These algorithms are trained using
chronological age as well as phenotypic ageing measures that consist of
clinical biomarkers, and are considered to represent the speed of bio­
logical age. Greater speed of ageing measured with these algorithms is
associated with an increased risk of mortality (Rutledge et al., 2022).
DNAm ageing scores reflecting older biological than chronological age
were found to be associated with a variety of social factors, such as so­
cioeconomic conditions (Fiorito et al., 2019; Hughes et al., 2018).
However, the associations with family factors have not been compre­
hensively examined.
The literature on social determinants of health and DNA methylation
is largely focused on old age groups or is limited in the number of age
groups examined (Evans et al., 2021). Here we examine the association
of partnership status with two DNAm age algorithms, ‘PhenoAge’ and
‘DunedinPACE’. Due to the processes of their derivation, we will refer to
"faster" DunedinPACE and "older" PhenoAge when considering biolog­
ical age that is greater than chronological age (‘accelerated’ age). As the
literature suggests that there may be sex interactions in the association
of partnership status and health (Wang et al., 2020), we explore in­
teractions with sex. Further, as partnership status may have different
saliency in younger and older age groups (Requena & Reher, 2021), we
explore interactions with age group. In addition to age and sex, a
number of factors are associated with partnership and dissolution that
could play a role in the association between partnership status and
biological age. For example, since there is evidence on the connection
between living arrangements, marital status (Grundström et al., 2021)
and mental health, household composition may represent a psychosocial
influence of the relationship between partnership status and biological
ageing. Besides, a large body of evidence suggests that marital status is
protective against poor health behaviours (Schone & Weinick, 1998).
Based on the previous evidence and the above theoretical frame­
work, we address the following research hypotheses.
Hypothesis 1. That being partnered is associated with negative
accelerated age, that is younger biological age compared to chronolog­
ical age, compared to single, divorced or widowed counterparts.
Hypothesis 2. The association between partnership status and age
acceleration differs between men and women. The impact of relation­
ship disruption on men’s biological age is greater than on women such
that accelerated age will be apparent in unpartnered men compared to
their partnered counterparts.
Hypothesis 3. The association between partnership status and age
acceleration differs according to life stage. Partnership status will have a
greater adverse association on older people’s biological age but not for
younger age groups.
2. Methods
2.1. Study design and study participants
Understanding Society, the UK Household Longitudinal Study
W. Wang et al. SSM - Population Health 24 (2023) 101551

(UKHLS), is a large, ongoing study designed to be representative of the civil partners, and data points greater than 3 standard deviations (SD)
UK population. We utilize information from the DNAm subsample from from DNAm age mean values, there were 3492 participants included in
Understanding Society. Between May 2010 and July 2012 (Waves 2 and the analysis (Fig. 1).
3), participants’ blood samples were collected during nurse visits in the
participants’ homes, which followed, on average, 5 months after a ‘main
2.2. Measures
study’ interview. Respondents were eligible for a Nurse visit if they were
living in England, Scotland or Wales, adult (aged 16 or older), and met
2.2.1. Independent variable: partnership status
other conditions detailed in the user’s guide (Understanding Society:
The exposure was self-reported current de facto marital status re­
Waves 2 and 3 Nurse Health Assessment, 2010–2012, 2022). All data
ported in the main survey, here referred to as “partnership status”.
collection methods were carried out in accordance with Information
Partnership status was categorized into 4 groups, including married or
Commissioner’s Office guidelines and regulations. Informed consent
living with a partner (reference), single, divorced or separated,
was obtained from all participants. The University of Essex Ethics
widowed.
Committee approved data collection on Understanding Society main
study. Approval for asking for consent and the collection of biosocial
2.2.2. Outcome variable: DNAm age acceleration
data by trained nurses was obtained from the National Research Ethics
Biological ageing rates are calculated using DNAm markers ‘CpG
Service (Understanding Society—UK Household Longitudinal Study: A
sites’ across the genome (Bell et al., 2019). We calculated DNAm age
Biosocial Component, Oxfordshire A REC, Reference: 10/H0604/2). The
using two DNAm algorithms: PhenoAge (Levine et al., 2018) and Dun­
DNAm sample was restricted to participants of white ethnicity who had
edinPACE (Belsky et al., 2022) as previously described. In UKHLS (Un­
consented to both blood sampling and genetic analysis. We used the
derstanding Society: Waves 2 and 3 Nurse Health Assessment,
Illumina Methylation EPIC BeadChip which integrated over 850,000
2010–2012, 2022), PhenoAge estimates the phenotypic age based on
methylation sites across the genome. After pre-processing quality con­
markers of tissue and immune function using 512 CpG sites in whole
trol steps, including outlier removal, filtering poor-quality probes and
blood, and DunedinPACE captures the rate of change across 19
quantile normalization, DNAm profiling data is available for 3654
health-related biomarkers using 172 CpG sites in whole blood. The
participants.
PhenoAge DNAm algorithm captures biological age while DunedinPACE
After excluding inapplicable and missing data on variables, same-sex
measures the biological ‘pace of ageing’, which is estimated as a

Fig. 1. Analytical sample selection process.

3
W. Wang et al.
measurement of an average rate of 1 year of biological ageing per year of
chronological ageing. DunedinPACE, is measured in years of accelerated
ageing per year of ageing. In order to measure age acceleration without
the effect of chronological age, we calculated ‘PhenoAge’ DNAm age
acceleration. PhenoAgeAA quantified by the difference between the
amount of ageing observed based on DNAm and the amount of ageing
expected based on chronological age, is calculated as the residual from
regressing DNAm age (dependent variable) on chronological age (in­
dependent variable). This study includes PhenoAgeAA and Dun­
edinPACE in the analysis. We also report the z-score standardized
PhenoAgeAA and DunedinPACE, which help compare biological age
metrics with "pace of ageing" metrics in descriptive analysis.
2.2.3. Covariates
After a literature review (Cheung & Mui, 2022; Fiorito et al., 2019;
Hughes et al., 2018; Mainous et al., 2011; Nevalainen et al., 2017; Ryan
et al., 2020; Schmitz et al., 2022; Yu, 2023), several important potential
covariates were identified:
Educational attainment. Highest educational qualification was cate­
gorized as no qualification, other qualification (entry level qualification
in the UK), General Certificate of Secondary Education (GCSE) etc. (level
1 and level 2 qualifications), General Certificate of Education Advanced
Level (A-level) etc. (level 3 qualifications), other higher degree (level 4
and level 5 qualifications), and degree (level 6 to 8 qualifications).
Household composition. It was categorized into living with others vs.
living alone. Psychological distress. The 12-item General Health Ques­
tionnaire (GHQ) was used to assess participant’s psychological distress
during the main survey. We used the Likert scoring method (0-1-2-3),
and the scores continuously ranged from 0 to 36, with higher scores
indicating greater distress. Since the GHQ score is non-normal distri­
bution, we used the mean GHQ score (11 in our sample) as a cut-off as
previously described (Shelton & Herrick, 2009) to create two groups: the
absence (GHQ score <11) and the presence (GHQ score ≥11) of psy­
chological distress. Living area. People’s living areas were self-reported
as living in urban or rural areas. Body Mass index. Height and weight
were measured by the nurse using a portable stadiometer and digital
floor scale (the Tanita BF 522 scales) (Understanding Society: Waves 2 and
3 Nurse Health Assessment, 2010–2012, 2022) to calculate body mass
index (BMI). Adiposity was indexed by BMI which was categorized as
underweight (<18.5), healthy-weight (18.5–24.9), overweight
(25.0–29.9), and obese (≥30.0). Smoking. Smoking status was collected
in the questionnaire in the main survey of wave 2 and the following
three categories were created: never smoker, ex-smoker and current
smoker. Alcohol consumption. Current drinking behaviour was assessed
by questionnaire in the main survey of wave 2. As drinking 6 or more
units in a single session is classed as binge drinking (Department of
Health and Social Care, 2021), the participant’s drinking intensity was
divided into none, 0–6 units (>0 and < 6), and 6+ units ( ≥ 6).
2.3. Statistical analysis
We compared the bivariate associations between covariates with
each exposure and outcomes, using adjusted Wald test to test overall
significance. To test for sex moderation, we created three Sex × Part­
nership Status interaction terms, one for each partnership status group
other than the reference group, and entered these as a set into the
regression equation predicting DNAmAA. The association between the
set of interaction terms and DNAmAA were statistically significant (P <
0.05), so all models were estimated separately by sex. Then we further
tested the rationale of additionally stratifying models by age group, the
p-value of the likelihood ratio test was also statistically significant.
Covariates were divided into different blocks based on the combi­
nation of existing literatures and Social Determinants of Health (SDH)
conceptual framework (Solar & Irwin, 2010). This study has 5 blocks of
covariates, including technical covariates (batch barcode and various
cell composition estimates (Houseman et al., 2012)), demographic
4
SSM - Population Health 24 (2023) 101551
variables (chronological age, age2, educational attainment), psychoso­
cial block (household composition and psychological distress), material
block (living area), and behavioural block (adiposity, smoking, and
alcohol consumption). We included the age-squared term in these ana­
lyses to account for potential nonlinearity in the association between age
and DNAmAA. Because of substantial data missingness, alcohol con­
sumption and psychological distress were examined in sensitivity anal­
ysis only.
Linear regression models were used to assess the association between
partnership status and DNAmAA. Firstly, we report models stratified by
sex. Model 1 estimates marital status differences in DNAmAA, adjusting
for the technical covariates and demographic variables. Models 2–4
separately adds the psychosocial block (Model 2), material block (Model
3) and behavioural block (Model 4) variables to Model 1. The fully
adjusted model (Model 5) examines the net association between part­
nership status and DNAmAA after controlling for all the covariates.
Then we report a fully adjusted linear regression model for the entire
sample (changing the age from the continuous variable to the categor­
ical variable), additionally stratified by age group (using participants’
age tertiles as cut-off points). In these models, partnership status cate­
gories were further collapsed into 3 groups (married or living with a
partner vs. single vs. divorced, separated, or widowed) to preserve suf­
ficient sample size since widowed subjects were too few in the youngest
age group to analyze as a separate category.
There were two different sensitivity analyses: one restricted the
lower age limit of participants in order to exclude the selection effect of
age on partnership status, and the other added alcohol consumption and
psychological distress to the fully adjusted linear regression model. In
the age-restricted analyses, we sequentially excluded those under the
UK’s legal minimum age (18 years) to enter a marriage (Model A1),
those normally have not graduated from university in UK (≤25 years)
(Model A2), and those under the mean average age of marriage at data
collection time (34 years for females and 36 years for males) (Average
age at marriage England and Wales, by gender | Statista) (Model A3).
All the analyses were performed with STATA 16 software.
3. Results
3.1. Description of the samples
Descriptive statistics for study variables are presented separately by
partnership status groups and sex in Table 1. Among men 76.35% were
partnered, 12.18% were single people, 7.88% divorced or separated,
and 3.59% widowed. The corresponding figures for women were
69.90%, 9.96%, 12.47%, 7.67%. Participants had a younger mean
DNAm age than chronological age in PhenoAge but not DunedinPACE.
For both DNAm algorithm, the mean biological age/average pace of
ageing per year for people in the single group was the youngest/slowest,
while that for the widowed group was the oldest/fastest.
3.2. Association of DNAmAA with partnership status by sex
We first evaluated multivariate association between partnership
status and DNAmAA after statistically adjusting for covariates and
stratifying by sex (Table 2). From Model 1 to Model 3, compared to
being partnered, being single was significantly associated with positive
DunedinPACE for females; being widowed showed significantly greater
positive DNAmAA for both algorithms; and being divorced or separated
was significantly associated with positive PhenoAgeAA for females and
positive DunedinPACE in men and women. Positive DunedinPACE in
single females was not robust to adjustment for behavioural variables
(model 4). When adjusting for all covariates (Model 5), widows were
epigenetically “older" in comparison with partnered people for both
algorithms. Widows’ PhenoAgeAA was 2.00 years older (95% CI = 0.46,
3.54) for men and 1.29 years older (95% CI = 0.25, 2.33) for women;
DunedinPACE was 0.05 biological years per chronological year faster
W. Wang et al. SSM - Population Health 24 (2023) 101551

Table 1
Characteristics of the analytical sample in study of DNA methylation age (n = 3492), UK household longitudinal study.
Sex Male Female

Partnership status Partnership status

N (%)/Mean (SD) N (%)/Mean (SD)

Variable Total Partnered (married Single Divorced or Widowed Total Partnered (married Single Divorced or Widowed
or living with a separated or living with a separated
partner) partner)

n = 1535 n = 1172 (76.35) n = 187 n = 121 n = 55 n = 1957 n = 1368 (69.90) n = 195 n = 244 n = 150
(12.18) (7.88) (3.59) (9.96) (12.47) (7.67)

Age (years) 52.68 53.90 (14.27) 37.63 56.18 (9.89) 70.23 51.98 51.84 (13.74) 36.17 54.37 69.91
(15.58) (17.03) (10.27) (15.14) (15.33) (10.93) (10.73)
PhenoAge 45.28 46.09 (11.48) 33.72 48.60 (8.56) 59.91 44.85 44.40 (10.87) 33.56 47.86 (9.75) 58.71
(years) (12.77) (15.48) (7.88) (12.03) (12.32) (9.22)
PhenoAgeAA − 0.05 − 0.12 (4.90) − 0.61 0.72 (5.71) 1.77 0.04 − 0.31 (5.17) 0.30 1.30 (5.72) 0.80
(5.08) (5.33) (6.04) (5.33) (5.28) (5.79)
PhenoAgeAA (z- − 0.01 − 0.02 (0.94) − 0.12 0.14 (1.09) 0.34 0.01 − 0.06 (0.99) 0.06 0.25 (1.10) 0.15
score) (0.97) (1.02) (1.16) (1.02) (1.01) (1.11)
DunedinPACE 1.07 1.06 (0.13) 1.02 1.12 (0.17) 1.16 1.05 1.04 (0.13) 1.05 1.08 (0.13) 1.12
(0.14) (0.14) (0.12) (0.13) (0.14) (0.13)
DunedinPACE 0.60 0.04 (0.97) − 0.26 0.46 (1.25) 0.79 − 0.05 − 0.14 (0.95) − 0.07 0.20 (0.95) 0.45
(z-score) (1.03) (1.06) (0.91) (0.98) (1.04) (0.96)
Highest educational qualification
Degree 367 300 (25.60) 38 23 (19.01) 6 (10.91) 391 301 (22.00) 40 38 (15.57) 12 (8.00)
(23.91) (20.32) (19.98) (20.51)
Other higher 179 142 (12.12) 17 (9.09) 12 (9.92) 8 (14.55) 292 214 (15.64) 25 32 (13.11) 21 (14.00)
degree (11.66) (14.92) (12.82)
A-level 347 263 (22.44) 52 24 (19.83) 8 (14.55) 312 220 (16.08) 47 36 (14.75) 9 (6.00)
(22.61) (27.81) (15.94) (24.10)
GCSE 327 238 (20.31) 47 26 (21.49) 16 (29.09) 470 319 (23.32) 59 64 (26.23) 28 (18.67)
(21.30) (25.13) (24.02) (30.26)
Other 152 112 (9.56) 17 (9.09) 18 (14.88) 5 (9.09) 209 133 (9.72) 14 (7.18) 30 (12.30) 32 (21.33)
qualification (9.90) (10.68)
No qualification 163 117 (9.98) 16 (8.56) 18 (14.88) 12 (21.82) 283 181 (13.23) 10 (5.13) 44 (18.03) 48 (32.00)
(10.62) (14.46)
Household composition
Living alone 233 6 (0.51) 87 96 (79.34) 44 (80.00) 301 6 (0.44) 67 111 (45.49) 117
(15.18) (46.52) (15.38) (34.36) (78.00)
Living with 1304 1166 (99.49) 100 25 (20.66) 11 (20.00) 1656 1362 (99.56) 128 133 (54.51) 33 (22.00)
others (84.82) (53.48) (84.62) (65.64)
Living area
Urban area 1085 811 (69.20) 146 87 (71.90) 42 (74.55) 1421 958 (70.03) 168 184 (75.41) 111
(70.68) (78.07) (72.61) (86.15) (74.00)
Rural area 450 361 (30.80) 41 34 (28.10) 14 (25.45) 536 410 (29.97) 27 60 (24.59) 39 (26.00)
(29.32) (21.93) (27.39) (13.85)
Adiposity
Healthy-weight 365 243 (20.73) 68 38 (31.40) 16 (29.09) 638 442 (32.31) 82 82 (33.61) 32 (21.33)
(23.78) (36.36) (32.60) (42.05)
Underweight 8 (0.52) 2 (0.17) 6 (3.21) 0 (0.00) 0 (0.00) 24 11 (0.80) 8 (4.10) 2 (0.82) 3 (2.00)
(1.23)
Overweight 676 538 (45.90) 72 50 (41.32) 16 (29.09) 692 496 (36.26) 59 77 (31.56) 60 (40.00)
(44.04) (38.50) (35.36) (30.26)
Obese 486 389 (33.19) 41 33 (27.27) 23 (41.82) 603 419 (30.63) 46 83 (34.02) 55 (36.67)
(31.66) (21.93) (30.81) (23.59)
Smoking
Never smoker 727 551 (47.01) 104 50 (41.32) 22 (40.00) 1085 798 (58.33) 103 104 (42.62) 80 (53.33)
(47.36) (55.61) (55.44) (52.82)
Ex-smoker 509 422 (36.01) 32 31 (25.62) 24 (43.64) 503 359 (26.24) 27 71 (29.10) 46 (30.67)
(33.16) (17.11) (25.70) (13.85)
Current smoker 299 199 (16.98) 51 40 (33.06) 9 (16.36) 369 211 (15.42) 65 69 (28.28) 24 (16.00)
(19.48) (27.27) (18.86) (33.33)

DNAm: DNAm methylation.

Adiposity was indexed by BMI which was categorized as underweight (<18.5), healthy-weight (18.5–24.9), overweight (25.0–29.9), and obese (≥30.0).

(95% CI = 0.01, 0.08) for men and 0.03 biological years per chrono­
logical year faster (95% CI = 0.01, 0.05) for women. Positive age ac­
celeration is apparent in all groups in final models in comparison to
partnered groups, however, these findings were not all significant.
3.3. Results when considering the age moderation effect
Next, we evaluated whether the association between marital status
and positive age acceleration changed when further stratified by age
groups (Table 3). In younger age groups, by the PhenoAge algorithm,
single men were epigenetically "younger" than chronological age (Phe­
noAgeAA was 1.96 years younger), and by DunedinPACE algorithm,
younger single men were ageing epigenetically slower (DunedinPACE
was 0.03 biological years per chronological year slower) compared to
partnered counterparts in the same age group. Singlehood was signifi­
cantly associated with greater epigenetic ageing (4.27 years older)
among older men, especially using the PhenoAge algorithm. Divorced,
separated, or widowed status was associated with greater epigenetic age

5
W. Wang et al. SSM - Population Health 24 (2023) 101551

Table 2
Regression models of the association between partnership status and DNAm AA, stratified by sex (males n = 1535, females n = 1957).
Model 1a Model 2b Model 3c Model 4d Model 5e

Male Female Male Female Male Female Male Female Male Female

Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95% Coeff. (95%
CI) CI) CI) CI) CI) CI) CI) CI) CI) CI)

Panel A: PhenoAgeAA
Partnership status
Partnered Reference
(married or
living with
a partner)
Single 0.31 (− 0.51 0.55 (− 0.25 0.30 (− 0.72 0.35 (− 0.52 0.30 (− 0.52 0.50 (− 0.30 0.38 (− 0.42 0.32 (− 0.46 0.29 (− 0.71 0.08 (− 0.79
to 1.13) to 1.35) to 1.32) to 1.23) to 1.12) to 1.30) to 1.18) to 1.11) to 1.29) to 0.94)
Divorced or 0.27 (− 0.62 1.14 0.26 (− 0.95 0.93 0.25 (− 0.63 1.11 0.25 (− 0.62 0.86 0.12 (− 1.07 0.63 (− 0.11
separated to 1.16) (0.48–1.79) to 1.47) (0.18–1.69) to 1.14) (0.45–1.77) to 1.12) (0.21–1.51) to 1.31) to 1.37)
*** * *** **
Widowed 2.23 1.89 2.22 1.57 2.19 1.85 2.15 1.65 2.00 1.29
(0.90–3.56) (1.00–2.78) (0.65–3.80) (0.51–2.63) (0.86–3.52) (0.96–2.75) (0.85–3.45) (0.77–2.52) (0.46–3.54) (0.25–2.33)
*** *** ** ** *** *** *** *** * *
R-squared 0.193 0.232 0.193 0.232 0.195 0.233 0.236 0.268 0.237 0.269
Adjusted R- 0.163 0.210 0.162 0.209 0.164 0.210 0.205 0.244 0.204 0.245
squared
Panel B: DunedinPACE
Partnership status
Partnered Reference
(married or
living with
a partner)
Single 0.01 (− 0.01 0.02 0.03 0.03 0.01 (− 0.01 0.02 0.01 (− 0.01 0.01 (− 0.01 0.02 (− 0.00 0.01 (− 0.01
to 0.03) (0.00–0.04) (0.00–0.05) (0.01–0.05) to 0.03) (0.00–0.04) to 0.03) to 0.03) to 0.05) to 0.03)
* * * *
Divorced or 0.04 0.03 0.06 0.03 0.04 0.03 0.03 0.02 0.04 0.02
separated (0.01–0.06) (0.01–0.05) (0.03–0.09) (0.01–0.05) (0.01–0.06) (0.01–0.05) (0.01–0.05) (0.00–0.03) (0.02–0.07) (0.00–0.03)
*** *** *** *** *** *** ** * ** *
Widowed 0.04 0.04 0.06 0.05 0.04 0.04 0.03 0.03 0.05 0.03
(0.01–0.08) (0.02–0.06) (0.02–0.10) (0.02–0.07) (0.01–0.07) (0.02–0.06) (0.00–0.06) (0.01–0.05) (0.01–0.08) (0.01–0.05)
* *** *** *** * *** * ** ** *
R-squared 0.322 0.228 0.342 0.228 0.326 0.230 0.455 0.402 0.457 0.403
Adjusted R- 0.297 0.205 0.298 0.205 0.299 0.207 0.432 0.383 0.434 0.383
squared

CI: Confidence interval.

Coef.: Coefficient.
* p < 0.05, ** p < 0.01, *** p < 0.001.
a
Adjusted for technical covariates (barcode and various cell composition estimates), demographic variables (chronological age, age2, education).
b
Adjusted for technical covariates, demographic variables, psychosocial block (household composition).
c
Adjusted for technical covariates, demographic variables, material block (living area).
d
Adjusted for technical covariates, demographic variables, behavioural block (adiposity, smoking).
e
Adjusted for technical covariates, demographic variables, psychosocial block, material block, behavioural block.

in younger (DunedinPACE was 0.09 biological years per chronological models, compared with partnered participants, DunedinPACE for
year faster) and older (DunedinPACE was 0.07 biological years per divorced/separated and widowed women was 0.02 (95% CI = 0.01,
chronological year faster) men compared to being partnered. 0.04) and 0.03 (95% CI = 0.01, 0.06) biological years per chronological
year faster, and PhenoAgeAA for divorced/separated and widowed
women was 1.09 (95% CI = 0.28, 1.90) and 1.63 (95% CI = 0.49, 2.76)
3.4. Sensitivity analysis
years older, respectively. The standardized regression coefficients shows
that partnership status has a strong association with biological ageing.
For sensitivity analysis (Table Appendix A), when excluding people
aged less than 18, less than 25, and less than UK average age at marriage,
4. Discussion
widows still exhibited positive DNAmAA in comparison with partnered
people, using both algorithms. Those divorced or separated males were
We demonstrate that partnership status is associated with DNAm age
ageing significantly faster than chronological age compared to part­
acceleration, such that both men and women who are widowed and
nered counterparts when using DunedinPACE algorithm. Among those
divorced are biologically older than chronological age to a greater
beyond the mean age at marriage in UK, the values for single males
extent than their counterparts in partnerships. We observe an interac­
measured by both algorithms became statistically significant, with
tion with sex and age; for example, being single for men was a protective
PhenoAgeAA 1.89 years older (95% CI = 0.43, 3.35) and DunedinPACE
factor against ageing at a younger age, but remaining single was asso­
0.05 biological years per chronological year faster (95% CI = 0.02, 0.08)
ciated with worse biological ageing with increasing age for both women
compared to partnered counterparts.
and men. Our results accord with previous observations that being in a
The addition of alcohol consumption and psychological distress to
relationship, including marriage or cohabitation, confers a health
the model reduced sample size (n = 3021) and hence power but did not
advantage (Crowley, 2019; Liu & Waite, 2014; McFarland et al., 2013;
affect conclusions in DunedinPACE for females (Table Appendix B).
Yu, 2023).
When adding two more covariates to fully adjusted linear regression

6
W. Wang et al. SSM - Population Health 24 (2023) 101551

Table 3
Results of linear regressions using DNAm AA as outcomes, stratified by age group and sex.
Male Female
a b b
Full sample (n 16–44 years 45–60 years 61 years and Full samplea (n 16–44 yearsb 45–60 yearsb 61 years and
= 1535) olderb = 1957) olderb

Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% Coeff. (95% Coeff. (95%
CI) CI) CI)

Panel A: PhenoAgeAA
Partnership status
Partnered (married or Reference Reference
living with a partner)
Single − 0.68 (− 1.60 to − 1.96 (-3.01 to 1.66 (− 0.39 to 4.27 (0.94 to − 0.23 (− 1.04 to − 0.24 (− 1.20 1.24 (− 0.59 − 1.90 (− 4.88
0.24) -0.91)*** 3.71) 7.60)** 0.58) to 0.72) to 3.08) to 1.08)
Divorced, separated, or 0.19 (− 0.92 to 1.13 (− 1.20 to 0.80 (− 1.07 to 2.34 (− 032 to 0.65 (− 0.05 to − 0.13 (− 1.52 0.56 (− 0.48 1.26 (− 0.63
widowed 1.30) 3.46) 2.66) 4.99) 1.35) to 1.26) to 1.59) to 3.15)
Panel B: DunedinPACE
Partnership status
Partnered (married or Reference Reference
living with a partner)
Single − 0.01 (− 0.03 to − 0.03 (− 0.06 to 0.05 0.07 (− 0.01 to 0.01 (− 0.01 to 0.01 (− 0.01 0.02 (− 0.02 0.01 (− 0.05
0.01) − 0.01)* (0.00–0.09)* 0.14) 0.03) to 0.03) to 0.06) to 0.07)
Divorced, separated, or 0.03 (0.00–0.05) 0.09 (0.03–0.14) 0.04 (− 0.01 to 0.07 0.02 (0.01–0.04) 0.00 (− 0.03 0.02 (− 0.00 0.03 (− 0.01
widowed * ** 0.08) (0.02–0.13)* ** to 0.04) to 0.04) to 0.07)

* p < 0.05, ** p < 0.01, *** p < 0.001.

a
Adjusted for technical covariates, age categories, education, household composition, living area, adiposity, smoking.
b
Adjusted for technical covariates, education, household composition, living area, adiposity, smoking.

We observe that divorced, separated, or widowed people age faster,
when assessed with DNA methylation, compared to counterparts in
partnerships, which is consistent with Hypothesis 1. Our observations
accord with studies that suggest that divorce and widowhood are asso­
ciated with an increased risk of poor health (Tatangelo et al., 2017).
Such findings may be explained by the stress model, which states that
relationship disruption may serve as a psychosocial stressor, and psy­
chosocial stress may be associated with DNAm changes (Palma-Gudiel
et al., 2015; Schiele et al., 2020). Our finding that both single older men
and women were biologically older than their counterparts accord with
the resource model, which proposes that being single may represent a
lack of material and social resources, with a resultant adverse associa­
tion with health.
Our findings do not fully accord with a recent study which found
greater epigenetic ageing among single women but not for single men in
the Health and Retirement Study (HRS) (Yu, 2023). Here, consistent
with hypothesis 2, we found that the association between partnership
status and DNAm age acceleration is more pronounced in men than
women, especially using DunedinPACE algorithm. Such findings may be
explained by previous research on observed patterns in sex differences in
social network composition, in which men typically receive more
emotional support and favourable regulation of health behaviours from
family than women do (Carr & Springer, 2010; McKenzie et al., 2018;
Williams & Umberson, 2004), thus men’s health status may be more
influenced by partnership status.
Age-related changes in biology can occur before middle age, how­
ever, the study of marital status and epigenetic markers of age has been
restricted to older age groups. Studies have examined the associations
between partnership status and health conditions in younger age groups,
but less using DNAm as an ageing biomarker like us. Interestingly,
Henning-Smith and Gonzales (2020) who also divided participants to
younger adults, middle-aged adults, and older adults, described a similar
pattern of association by age of marital status and self-rated health as we
observed with ‘PhenoAge’ (Henning-Smith & Gonzales, 2020). The
literature on additional molecular biomarkers of ageing, such as telo­
mere length and attrition (Mainous et al., 2011; Chen et al., 2020;
Whisman et al., 2016; Yu & Liu, 2021) are also limited to middle and
older age groups. For example Mainous (2011) described the association
between marital status and telomere length in groups aged 40–64, and
reported shorter telomere length in unmarried groups independently of
social support and health behaviours (Mainous et al., 2011). Therefore,
it is recommended that further studies explore the impact of partnership
status on DNA methylation algorithms and other biological ageing
measurements from young adults and for the entire age groups. In
addition, there have been secular changes in the rates of marriage,
informal partnerships, divorce and age of first marriage. Studies that
have restricted age groups are likely not be able to observe the impacts
of these changes. Thus, our study is likely to capture this greater rate of
informal partnership in younger or middle-aged participants not
apparent in studies that are restricted to older age groups. Indeed, we
observe that our sample has more people in partnerships in middle age,
while we observe the greatest rate of widowhood in our older age
groups.
We found single status is sensitive to ageing and differs between men
and women, which is partially consistent with Hypothesis 2 and 3. Our
study found that at a younger age, partnered men were biologically
older than non-partnered counterparts. This finding suggests that
"adverse selection" may be at play in young men, i.e. less healthy people
may be more inclined to enter into marriage, which is consistent with
previous research (Mr & Sloggett, 1998). This contradicts the literature
on “positive selection” (Murray, 2000), which suggests that more
healthy people enter into partnerships.
Our finding that partnered older age men are biologically younger
may provide insight into the association of partnership dissolution and
biological age as it is likely that those enter an early relationship may
also be more likely to also dissolute relationships more readily. An
alternative explanation is there is reverse causation such that those in
poor health may not marry in middle age. We couldn’t address selection
out of marriage directly as we conducted a cross sectional analysis.
Although our age specific analyses suggest that unmarried younger men
are healthier than married counterparts arguing for "adverse selection"
explanation, these may represent cohort specific effects and therefore
may not provide insight into mechanisms that may be in play for older
age groups. For example, we did not know whether healthy people may
be able to opt out of marriage more freely, which we have not accounted
for in our analyses. For all DNAm algorithms, we found being widowed
was associated with DNAmAA in both sexes. At a population level, given
that women generally have high probability of outliving men (Zarulli
et al., 2018), they may be more likely than men to become widowed and
remain widowed longer in their later lives.

7
W. Wang et al.
A number of potential pathways are proposed in our study, including
material, psychosocial and behavioural pathways. The biomarkers of
ageing employed in this work are associated with a number of factors.
Many associations with covariates have been reported previously, for
example disadvantaged SEP and positive accelerated age (Bao et al.,
2022; Evans et al., 2021; George et al., 2021; Hughes et al., 2018), but
others, such as the positive accelerated age we observe in groups living
in urban areas, which have not been described previously (Evans et al.,
2021) may require further investigation.
We infer that biological ageing processes underly some psychosocial,
material, and behavioural mechanisms, as we see associations between
partnership status and DNAm independently and with a number of ad­
justments. We did not see an association between our ageing biomarkers
and psychological distress, but this may be because our instrument, the
GHQ, may not be sensitive to the measures of mental health pertinent to
marital status or that we have not examined the psychosocial environ­
ment more broadly, for example marital relationship quality, which is
associated with health (Kiecolt-Glaser & Wilson, 2017). Although this
study controlled for household composition and psychological distress
as psychosocial factors, future research should consider other
stress-related psychosocial factors. For example, stress biomarkers such
as cortisol (Adam & Kumari, 2009), inflammatory markers (Marsland
et al., 2017) or allostatic load (Guidi et al., 2021) could be controlled in
order to investigate whether the association between partnership status
and epigenetic ageing rate is mediated through stress-related pathways.
This study had several strengths as it is based on a national study, it
comprised a large sample with representation from almost the entire
adult age range. We examined more than 1 s generation algorithm.
However, the sample was restricted to those reporting white/European
ethnicity, and therefore results may not be generalizable to other ethnic
groups. Measurement of DNAm was made on one occasion in adults and
therefore it is not possible to examine whether age acceleration occurred
before divorce or widowhood or vice versa and further it is not possible
to assess within-person change in adulthood. We did not examine
quality of marital relationships, which are also associated with health
(Kiecolt-Glaser & Wilson, 2017) and may precede marital dissolution.
Our analyses suggest that psychological distress, which may be associ­
ated with poor marital relationships, divorce or widowhood (Foran
et al., 2015) did not play a role in the associations examined. However,
there was a five-month delay between the collection of mental health
data and blood sample collection and we do not know the stability or not
of the DNAmethylation algorithms. The material, psychosocial and
behavioural factors examined in this study may be mediators rather than
moderators or confounders and thus there may be overadjustment. We
examined two algorithms - there are additional algorithms that have
been developed and can be tested. We did not adjust for multiple testing,
as the strength of associations indicated that our main conclusions
would not be altered by a Bonferroni correction. In stratified analyses,
some numbers were small and therefore it is possible that these analyses
are underpowered. The algorithms employed in this analysis have been
tested and trained against biomarkers that vary with age and may
therefore reflect the processes that they capture. The variety of associ­
ations apparent for these algorithms in this and additional studies (Bao
et al., 2022; Raffington & Belsky, 2022; Thomas et al., 2023) suggests
that our analyses may be subject to residual confounding. A focus on
estimators of age restricts analyses to a small subset of methylation sites
across the genome and thus limits an analysis of partnership status with
DNA methylation more broadly. Broader analyses have the potential to
uncover wider pathways by which social factors and health might be
associated.
8
SSM - Population Health 24 (2023) 101551
5. Conclusions
We conclude that the health disadvantage associated with divorce/
separation and widowhood is apparent when examining biomarkers of
ageing measured with DNA methylation across the adult age span. For
single people, there is heterogeneity in our observations such that age
and sex specific associations are apparent. The findings highlight the
importance of partnership in people’s ageing process, and these findings
will be helpful for health policymakers and practitioners who seek to
better design effective interventions targeted at vulnerable sub­
populations and life stages to minimize inequalities in the biological
ageing process.
Ethical statement
All methods were carried out in accordance with Information Com­
missioner’s Office guidelines and regulations. Informed consent was
obtained from all participants. The University of Essex Ethics Committee
approved data collection on Understanding Society main study. Approval
for asking for consent and the collection of biosocial data by trained
nurses was obtained from the National Research Ethics Service (Un­
derstanding Society—UK Household Longitudinal Study: A Biosocial
Component, Oxfordshire A REC, Reference: 10/H0604/2).
Funding
DNA methylation measurement in UKHLS was funded through en­
hancements to the Economic and Social Research Council (ESRC) Grants
ES/K005146/1 and ES/N00812X/1. WW and AD are Soc-B students
(ESRC ES/T00200X/1, project reference numbers: 2765592 and
2765580 respectively). Y.B. was partially supported by the ESRC (ES/
N00812X/1). M.K. is supported by the University of Essex, ESRC (RES-
596-28-0001) and ESRC (ES/S012486/1; ES/T014083/1).
CRediT authorship contribution statement
Wen Wang: Formal analysis, investigation, writing – original draft,
writing – review & editing.
Anna Dearman: Writing – review & editing.
Yanchun Bao: conceptualisation, investigation, writing – review &
editing.
Meena Kumari: conceptualisation, supervision, writing – review &
editing.
Declaration of competing interest
The authors declare no competing interests.
Data availability
All data are available from the UKDA. UK Data Service. SN:7251.
10.5255/UKDA-SN-7251-3
Acknowledgements
Understanding Society is an initiative funded by the Economic and
Social Research Council and various Government Departments, with
scientific leadership by the Institute for Social and Economic Research,
University of Essex, and survey delivery by NatCen Social Research and
Kantar Public. The research data are distributed by the UK Data Service.
W. Wang et al. SSM - Population Health 24 (2023) 101551

Appendix

Appendix A Results of sensitivity analyses controlling the age range of participants for regression models of the association between partnership status and
DNAm age acceleration, stratified by sex

Sensitivity analyses A1a Sensitivity analyses A2b Sensitivity analyses A3c

Male Female Male Female Male Female

Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI)

Panel A: PhenoAgeAA
Partnership status
Partnered (married or living with a Reference
partner)
Single 0.38 (− 0.64 0.18 (− 0.69 to 0.91 (− 0.30 to 0.01 (− 0.95 to 1.89 (0.43 to 3.35) − 0.15 (− 1.32 to
to1.40) 1.06) 2.12) 0.96) * 1.01)
Divorced or separated 0.19 (− 1.01 to 0.67 (− 0.08 to 0.54 (− 0.75 to 0.62 (− 0.14 to 0.93 (− 0.49 to 0.65 (− 0.15 to 1.45)
1.39) 1.41) 1.83) 1.38) 2.35)
Widowed 2.04 (0.49–3.58)** 1.31 (0.27–2.35)* 2.31 (0.71–3.92)** 1.22 (0.16–2.27)* 2.60 (0.89–4.32)** 1.14 (0.04–2.24)*
Panel B: DunedinPACE
Partnership status
Partnered (married or living with a Reference
partner)
Single 0.02 (0.00–0.05)* 0.01 (− 0.01 to 0.03 (0.00–0.06)* 0.02 (− 0.01 to 0.05 (0.02–0.08)** 0.01 (− 0.01 to 0.04)
0.03) 0.04)
Divorced or separated 0.04 (0.02–0.07) 0.02 (0.00–0.03)* 0.05 (0.02–0.08)** 0.02 (0.00–0.03)* 0.06 (0.03–0.09) 0.02 (− 0.01 to 0.03)
*** ***
Widowed 0.05 (0.01–0.08)** 0.03 (0.01–0.05)* 0.05 (0.01–0.09)** 0.03 (0.00–0.05)* 0.06 (0.02–0.10)** 0.02 (− 0.01 to 0.05)
Adjusted for technical covariates, demographic variables, psychosocial block, material block, behavioural block.
* p < 0.05, ** p < 0.01, *** p < 0.001.
a
Excluded those under the legal minimum age (18 years) to enter into a marriage, n = 3460.
b
Keep those beyond 25 years old, n = 3334.
c
Keep those beyond the mean average age at marriage (34 years for female and 36 years for male), n = 2955.

Appendix B Results of sensitivity analyses adding drinking intensity and psychological distress to fully adjusted linear regressions model (n = 3021)

PhenoAgeAA z_PhenoAgeAA† DunedinPACE z_DunedinPACE†

Male Female Male Female Male Female Male Female

Coeff. (95% Coeff. (95% CI) Beta (95% CI) Beta (95% CI) Coeff. (95% CI) Coeff. (95% CI) Beta (95% CI) Beta (95% CI)
CI)

Partnership status
Partnered (married or Reference
living with a
partner)
Single 0.32 (− 0.78 to 0.33 (− 0.62 to 0.06 (− 0.15 to 0.06 (− 0.12 to 0.01 (− 0.01 to 0.01 (− 0.01 to 0.10 (− 0.09 to 0.09 (− 0.07 to
1.43) 1.28) 0.27) 0.24) 0.04) 0.03) 0.28) 0.24)
Divorced or separated 0.47 (− 0.83 to 1.09 (0.28 to 0.09 (− 0.16 to 0.21 (0.05 to 0.04 (0.02 to 0.02 (0.01 to 0.33 (0.11 to 0.15 (0.01 to
1.78) 1.90)** 0.34) 0.36)** 0.07)** 0.04)* 0.55)** 0.28)*
Widowed 1.78 1.63 0.34 0.31 0.03 (− 0.01 to 0.03 0.22 (− 0.06 to 0.26
(0.10–3.47)* (0.49–2.76)** (0.02–0.66)* (0.09–0.53)** 0.07) (0.01–0.06)** 0.51) (0.07–0.45)**
Age 0.33 0.23 0.96 0.67 0.01 − 0.01 (− 0.01 0.65 − 0.02 (− 0.27
(0.22–0.44) (0.13–0.33)*** (0.64–1.28) (0.37–0.96)*** (0.00–0.01)*** to 0.00) (0.37–0.93)*** to 0.23)
*** ***
Highest educational qualification
No qualification Reference
Other qualification − 0.76 (− 1.90 − 0.25 (− 1.20 − 0.14 (− 0.36 − 0.05 (− 0.23 − 0.04 (− 0.07 − 0.02 (− 0.04 − 0.29 (− 0.49 − 0.16 (− 0.32
to 0.38) to 0.69) to 0.07) to 0.13) to − 0.01)** to − 0.01)* to − 0.10)** to − 0.01)*
GCSE − 0.74 (− 1.72 − 0.97 (− 1.79 − 0.14 (− 0.33 − 0.19 (− 0.34 − 0.06 (− 0.08 − 0.03 (− 0.05 − 0.42 (− 0.58 − 0.23 (− 0.36
to 0.25) to − 0.15)* to 0.05) to − 0.03)* to − 0.03)*** to − 0.01)*** to − 0.25)*** to − 0.09)***
A-level − 0.80 (− 1.77 − 1.25 (− 2.16 − 0.15 (− 0.34 − 0.24 (− 0.41 − 0.05 (− 0.08 − 0.03 (− 0.05 − 0.40 (− 0.57 − 0.24 (− 0.39
to 0.16) to − 0.35)** to 0.03) to − 0.07)** to − 0.03)*** to − 0.01)** to − 0.24)*** to − 0.09)**
Other higher degree − 0.95 (− 2.04 − 1.24 (− 2.14 − 0.18 (− 0.39 − 0.24 (− 0.41 − 0.06 (− 0.08 − 0.04 (− 0.06 − 0.42 (− 0.60 − 0.30 (− 0.45
to 0.14) to − 0.34)** to 0.03) to − 0.07)** to − 0.03)*** to − 0.02)*** to − 0.23)*** to − 0.15)***
Degree − 0.62 (− 1.61 − 1.05 (− 1.95 − 0.12 (− 0.31 − 0.20 (− 0.37 − 0.07 (− 0.09 − 0.05 (− 0.07 − 0.52 (− 0.69 − 0.35 (− 0.50
to 0.38) to − 0.15)* to 0.07) to − 0.03)* to − 0.05)*** to − 0.03)*** to − 0.35)*** to − 0.20)***
Household composition
Living alone Reference
Living with others 0.17 (− 0.99 to − 0.05 (− 0.94 0.03 (− 0.19 to − 0.01 (− 0.18 0.01 (− 0.01 to 0.01 (− 0.01 to 0.11 (− 0.09 to 0.05 (− 0.09 to
1.34) to 0.84) 0.26) to 0.16) 0.04) 0.03) 0.30) 0.20)
Living area
Urban area Reference
Rural area − 0.21 (− 0.76 − 0.14 (− 0.65 − 0.04 (− 0.14 − 0.03 (− 0.13 − 0.01 (− 0.02 − 0.01 (− 0.02 − 0.07 (− 0.16 − 0.06 (− 0.15
to 0.34) to 0.37) to 0.06) to 0.07) to 0.00) to 0.01) to 0.02) to 0.02)
(continued on next page)

9
W. Wang et al. SSM - Population Health 24 (2023) 101551

(continued )
PhenoAgeAA z_PhenoAgeAA† DunedinPACE z_DunedinPACE†

Male Female Male Female Male Female Male Female

Coeff. (95% Coeff. (95% CI) Beta (95% CI) Beta (95% CI) Coeff. (95% CI) Coeff. (95% CI) Beta (95% CI) Beta (95% CI)
CI)

Smoking
Never smoker Reference
Ex-smoker 1.08 0.46 (− 0.08 to 0.21 0.09 (− 0.02 to 0.03 0.04 0.21 0.28
(0.49–1.66) 1.00) (0.09–0.32) 0.19) (0.01–0.04)*** (0.03–0.05)*** (0.11–0.31)*** (0.19–0.37)***
*** ***
Current smoker 2.37 2.14 0.45 0.41 0.12 0.13 0.93 0.94
(1.66–3.08) (1.49–2.78)*** (0.32–0.59) (0.29–0.53)*** (0.11–0.14)*** (0.11–0.14)*** (0.81–1.05)*** (0.83–1.04)***
*** ***
Adiposity
Healthy-weight Reference
Underweight 0.47 (− 3.05 to − 1.54 (− 3.71 0.09 (− 0.58 to − 0.30 (− 0.71 − 0.02 (− 0.10 − 0.05 (− 0.10 − 0.17 (− 0.77 − 0.36 (− 0.71
3.99) to 0.62) 0.76) to 0.12) to 0.06) to 0.00) to 0.42) to 0.01)
Overweight 1.13 0.01 (− 0.54 to 0.22 0.01 (− 0.10 to 0.02 0.02 0.17 0.18
(0.47–1.78) 0.56) (0.09–0.34) 0.11) (0.01–0.04)** (0.01–0.04)*** (0.06–0.28)** (0.09–0.27)***
*** ***
Obese 1.85 1.38 0.35 0.26 0.06 0.08 0.46 0.60
(1.15–2.55) (0.80–1.96)*** (0.22–0.49) (0.15–0.38)*** (0.05–0.08)*** (0.07–0.09)*** (0.34–0.58)*** (0.51–0.70)***
*** ***
Alcohol (drinking intensity in a single session)
None Reference
<6 units − 0.24 (− 1.00 − 0.07 (− 0.62 − 0.05 (− 0.19 − 0.01 (− 0.12 − 0.01 (− 0.03 − 0.01 (− 0.01 − 0.07 (− 0.20 − 0.02 (− 0.11
to 0.52) to 0.49) to 0.10) to 0.09) to 0.01) to 0.01) to 0.06) to 0.07)
6 units and more 0.35 (− 0.34 to 0.36 (− 0.22 to 0.07 (− 0.06 to 0.07 (− 0.04 to − 0.01 (− 0.02 − 0.01 (− 0.02 − 0.06 (− 0.18 − 0.04 (− 0.14
1.04) 0.95) 0.20) 0.18) to 0.01) to 0.01) to 0.06) to 0.06)
Psychological distress
Absence of Reference
psychological
distress
Presence of 0.22 (− 0.30 to 0.14 (− 0.32 to 0.04 (− 0.06 to 0.03 (− 0.06 to 0.01 (− 0.01 to 0.01 (− 0.01 to 0.06 (− 0.03 to 0.05 (− 0.02 to
psychological 0.74) 0.60) 0.14) 0.11) 0.02) 0.02) 0.15) 0.13)
distress
* p < 0.05, ** p < 0.01, *** p < 0.001.
†For easier comparison, the effect-sizes are reported as standardized regression coefficients.

References Chen, R., Zhan, Y., Pedersen, N., Fall, K., Valdimarsdóttir, U. A., Hägg, S., & Fang, F.
(2020). Marital status, telomere length and cardiovascular disease risk in a Swedish
prospective cohort. Heart, 106(4), 267–272. https://doi.org/10.1136/heartjnl-2019-
Adam, E. K., & Kumari, M. (2009). Assessing salivary cortisol in large-scale,
315629
epidemiological research. Psychoneuroendocrinology, 34, 1423–1436. https://doi.
Cheung, E. S. L., & Mui, A. C. (2022). Do home and community environments explain
org/10.1016/j.psyneuen.2009.06.011
self-rated health among older Canadians? Evidence from the 2018 Canadian housing
Bao, Y., Gorrie-Stone, T., Hannon, E., Hughes, A., Andrayas, A., Neilson, G., Burrage, J.,
survey. HERD: Health Environments Research & Design Journal, 15, 112–125. https://
Mill, J., Schalkwyk, L., & Kumari, M. (2022). Social mobility across the lifecourse
doi.org/10.1177/19375867221085603
and DNA methylation age acceleration in adults in the UK. Scientific Reports, 12,
Crowley, J. E. (2019). Does everything fall apart? Life assessments following a gray
Article 22284. https://doi.org/10.1038/s41598-022-26433-2
divorce. Journal of Family Issues, 40(11), 1438–1461. https://doi.org/10.1177/
Alcohol consumption and telomere length: Mendelian randomization clarifies alcohol’s
0192513X19839735
effects. Molecular Psychiatry, 27, (2022), 4001–4008. https://doi.org/10.1038/
Department of Health and Social Care. (2021). (Chapter 12): Alcohol. https://www.gov.
s41380-022-01690-9
uk/government/publications/delivering-better-oral-health-an-evidence-based-too
Bell, C. G., Lowe, R., Adams, P. D., Baccarelli, A. A., Beck, S., Bell, J. T.,
lkit-for-prevention/chapter-12-alcohol.
Christensen, B. C., Gladyshev, V. N., Heijmans, B. T., Horvath, S., Ideker, T.,
Effect of alcohol use disorder on cellular aging. Psychopharmacology, 236, (2019),
Issa, J. J., Kelsey, K. T., Marioni, R. E., Reik, W., Relton, C. L., Schalkwyk, L. C.,
3245–3255. https://doi.org/10.1007/s00213-019-05281-5
Teschendorff, A. E., Wagner, W., Zhang, K., & Rakyan, V. K. (2019). DNA
Ennis, J., & Majid, U. (2021). “Death from a broken heart”: A systematic review of the
methylation aging clocks: Challenges and recommendations. Genome Biology, 20,
relationship between spousal bereavement and physical and physiological health
249. https://doi.org/10.1186/s13059-019-1824-y
outcomes. Death Studies, 45, 538–551. https://doi.org/10.1080/
Belsky, D. W., Caspi, A., Arseneault, L., Baccarelli, A., Corcoran, D. L., Gao, X.,
07481187.2019.1661884
Hannon, E., Harrington, H. L., Rasmussen, L. J., Houts, R., Huffman, K., Kraus, W. E.,
Evans, L., Engelman, M., Mikulas, A., & Malecki, K. (2021). How are social determinants
Kwon, D., Mill, J., Pieper, C. F., Prinz, J. A., Poulton, R., Schwartz, J., Sugden, K.,
of health integrated into epigenetic research? A systematic review. Social Science &
et al. (2020). Quantification of the pace of biological aging in humans through a
Medicine, 273, Article 113738. https://doi.org/10.1016/j.socscimed.2021.113738
blood test, the DunedinPoAm DNA methylation algorithm. Elife, 9, Article e54870.
Fiorito, G., McCrory, C., Robinson, O., Carmeli, C., Ochoa-Rosales, C., Zhang, Y., , …
https://doi.org/10.7554/eLife.54870
Matullo, G., et al. (2019). Socioeconomic position, lifestyle habits and biomarkers of
Belsky, D. W., Caspi, A., Corcoran, D. L., Sugden, K., Poulton, R., Arseneault, L.,
epigenetic aging: A multi-cohort analysis. Aging (Albany NY), 11(7), 2045–2070.
Baccarelli, A., Chamarti, K., Gao, X., Hannon, E., Harrington, H. L., Houts, R.,
https://doi.org/10.18632/aging.101900
Kothari, M., Kwon, D., Mill, J., Schwartz, J., Vokonas, P., Wang, C., Williams, B. S., &
Foran, H. M., Whisman, M. A., & Beach, S. R. H. (2015). Intimate partner relationship
Moffitt, T. E. (2022). DunedinPACE, a DNA methylation biomarker of the pace of
distress in the DSM-5. Family Process, 54(1), 48–63. https://doi.org/10.1111/
aging. Elife, 11, Article e73420. https://doi.org/10.7554/eLife.73420
famp.12122
Brenn, T., & Ytterstad, E. (2016). Increased risk of death immediately after losing a
Grundström, J., Konttinen, H., Berg, N., & Kiviruusu, O. (2021). Associations between
spouse: Cause-specific mortality following widowhood in Norway. Preventive
relationship status and mental well-being in different life phases from young to
Medicine, 89, 251–256. https://doi.org/10.1016/j.ypmed.2016.06.019
middle adulthood. SSM - Population Health, 14, Article 100774. https://doi.org/
Bulanda, J. R., Brown, J. S., & Yamashita, T. (2016). Marital quality, marital dissolution,
10.1016/j.ssmph.2021.100774
and mortality risk during the later life course. Social Science & Medicine, 165,
Guidi, J., Lucente, M., Sonino, N., & Fava, G. A. (2021). Allostatic load and its impact on
119–127. https://doi.org/10.1016/j.socscimed.2016.07.025
health: A systematic review. Psychotherapy and Psychosomatics, 90, 11–27. https://
Carr, D., & Springer, K. W. (2010). Advances in families and health research in the 21st
doi.org/10.1159/000510696
century. Journal of Marriage and Family, 72, 743–761. https://doi.org/10.1111/
Henning-Smith, C., & Gonzales, G. (2020). The relationship between living alone and
j.1741-3737.2010.00728.x
self-rated health varies by age: Evidence from the national health interview survey.

10
W. Wang et al.
Journal of Applied Gerontology, 39(9), 971–980. https://doi.org/10.1177/
0733464819835113
Houseman, E. A., Accomando, W. P., Koestler, D. C., Christensen, B. C., Marsit, C. J.,
Nelson, H. H., Wiencke, J. K., & Kelsey, K. T. (2012). DNA methylation arrays as
surrogate measures of cell mixture distribution. BMC Bioinformatics, 13(1), 1–6.
https://doi.org/10.1186/1471-2105-13-86
Hughes, A., Smart, M., Gorrie-Stone, T., Hannon, E., Mill, J., Bao, Y., Burrage, J.,
Schalkwyk, L., & Kumari, M. (2018). Socioeconomic position and DNA methylation
age acceleration across the life course. American Journal of Epidemiology, 187(11),
2346–2354. https://doi.org/10.1093/aje/kwy155
Hughes, M. E., & Waite, L. J. (2009). Marital biography and health at mid-life. Journal of
Health and Social Behavior, 50, 344–358. https://doi.org/10.1177/
002214650905000307
Kalmijn, M. (2003). Shared friendship networks and the life course: An analysis of survey
data on married and cohabiting couples. Social Networks, 25(3), 231–249. https://
doi.org/10.1016/S0378-8733(03)00010-8
Kalmijn, M. (2012). Longitudinal analyses of the effects of age, marriage, and parenthood
on social contacts and support. Advances in Life Course Research, 17(4), 177–190.
Longitudinal analyses of the effects of age, marriage, and parenthood on social
contacts and support.
Kiecolt-Glaser, J. K. (2018). Marriage, divorce, and the immune system. American
Psychologist, 73(9), 1098–1108. https://doi.org/10.1037/amp0000388
Kiecolt-Glaser, J. K., & Wilson, S. J. (2017). Lovesick: How couples’ relationships
influence health. Annual Review of Clinical Psychology, 13, 421. https://doi.org/
10.1146/annurev-clinpsy-032816-045111
Kojima, G., Walters, K., Iliffe, S., Taniguchi, Y., & Tamiya, N. (2020). Marital status and
risk of physical frailty: A systematic review and meta-analysis. Journal of the
American Medical Directors Association, 21(3), 322–330. https://doi.org/10.1016/j.
jamda.2019.09.017
Leopold, T. (2018). Gender differences in the consequences of divorce: A study of
multiple outcomes. Demography, 55(3), 769–797. https://doi.org/10.1007/s13524-
018-0667-6
Levine, M. E., Lu, A. T., Quach, A., Chen, B. H., Assimes, T. L., Bandinelli, S., Hou, L.,
Baccarelli, A. A., Stewart, J. D., Li, Y., Whitsel, E. A., Wilson, J. G., Reiner, A. P.,
Aviv, A., Lohman, K., Liu, Y., Ferrucci, L., & &Horvath, S. (2018). An epigenetic
biomarker of aging for lifespan and healthspan. Aging (Albany NY), 10(4), 573–591.
https://doi.org/10.18632/aging.101414
Liu, H., & Waite, L. (2014). Bad marriage, broken heart? Age and gender differences in
the link between marital quality and cardiovascular risks among older adults.
Journal of Health and Social Behavior, 55(4), 403–423. https://doi.org/10.1177/
0022146514556893
Liu, H., Zhang, Z., Choi, S.-w., & Langa, K. M. (2020). Marital status and dementia:
Evidence from the health and retirement study. The Journals of Gerontology: Serie
Bibliographique, 75, 1783–1795. https://doi.org/10.1093/geronb/gbz087
López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M., & Kroemer, G. (2013). The
hallmarks of aging. Cell, 153(6), 1194–1217. https://doi.org/10.1016/j.
cell.2013.05.039
Mainous, A. G., III, Everett, C. J., Diaz, V. A., Baker, R., Mangino, M., Codd, V., &
Samani, N. J. (2011). Leukocyte telomere length and marital status among middle-
aged adults. Age and Ageing, 40(1), 73–78. https://doi.org/10.1093/ageing/afq118
Marsland, A. L., Walsh, C., Lockwood, K., & John-Henderson, N. A. (2017). The effects of
acute psychological stress on circulating and stimulated inflammatory markers: A
systematic review and meta-analysis. Brain, Behavior, and Immunity, 64, 208–219.
https://doi.org/10.1016/j.bbi.2017.01.011
McFarland, M. J., Hayward, M. D., & Brown, D. (2013). I’ve got you under my skin:
Marital biography and biological risk. Journal of Marriage and Family, 75, 363–380.
https://doi.org/10.1111/jomf.12015
McKenzie, S. K., Collings, S., Jenkin, G., & River, J. (2018). Masculinity, social
connectedness, and mental health: Men’s diverse patterns of practice. American
Journal of Men’s Health, 12, 1247–1261. https://doi.org/10.1177/
1557988318772732
Moon, J. R., Kondo, N., Glymour, M. M., & Subramanian, S. V. (2011). Widowhood and
mortality: A meta-analysis. PLoS One, 6(8), Article e23465. https://doi.org/
10.1371/journal.pone.0023465
Mr, Y. C., & Sloggett, A. (1998). Health and adverse selection into marriage: Evidence
from a study of the 1958 British birth cohort. Public Health, 112, 309–311.
Murray, J. E. (2000). Marital protection and marital selection: Evidence from a
historical-prospective sample of American men. Demography, 37(4). https://doi.org/
10.1353/dem.2000.0010
Obesity accelerates epigenetic aging in middle-aged but not in elderly individuals.
Clinical Epigenetics, 9, (2017), 1–9. https://doi.org/10.1186/s13148-016-0301-7
Palma-Gudiel, H., Córdova-Palomera, A., Leza, J. C., & Fañanás, L. (2015).
Glucocorticoid receptor gene (NR3C1) methylation processes as mediators of early
adversity in stress-related disorders causality: A critical review. Neuroscience &
Biobehavioral Reviews, 55, 520–535. https://doi.org/10.1016/j.
neubiorev.2015.05.016
Percheski, C., & Meyer, J. M. (2018). Health and union dissolution among parenting
couples: Differences by gender and marital status. Journal of Health and Social
Behavior, 59(4), 569–584. https://doi.org/10.1177/0022146518808707
11
SSM - Population Health 24 (2023) 101551
Raffington, L., & Belsky, D. W. (2022). Integrating DNA methylation measures of
biological aging into social determinants of health research. Current Environmental
Health Reports, 9(2), 196–210. https://doi.org/10.1007/s40572-022-00338-8
Requena, M., & Reher, D. (2021). Partnership and mortality in mid and late life:
Protection or selection? Social Science & Medicine, 279, Article 113971. https://doi.
org/10.1016/j.socscimed.2021.113971
Rote, S. (2017). Marital disruption and allostatic load in late life. Journal of Aging and
Health, 29(4), 688–707. https://doi.org/10.1177/0898264316641084
Rutledge, J., Oh, H., & Wyss-Coray, T. (2022). Measuring biological age using omics
data. Nature Reviews Genetics, 23(22), 715–727. https://doi.org/10.1038/s41576-
022-00511-7
Ryan, J., Wrigglesworth, J., Loong, J., Fransquet, P. D., & Woods, R. L. (2020).
A systematic review and meta-analysis of environmental, lifestyle, and health factors
associated with DNA methylation age. The Journals of Gerontology: Series A, 75,
481–494. https://doi.org/10.1093/gerona/glz099
Schiele, M. A., Gottschalk, M. G., & Domschke, K. (2020). The applied implications of
epigenetics in anxiety, affective and stress-related disorders-A review and synthesis
on psychosocial stress, psychotherapy and prevention. Clinical Psychology Review, 77,
Article 101830. https://doi.org/10.1016/j.cpr.2020.101830
Schone, B. S., & Weinick, R. M. (1998). Health-related behaviors and the benefits of
marriage for elderly persons. The Gerontologist, 38(5), 618–627. https://doi.org/
10.1093/geront/38.5.618
Schreier, H., & Chen, E. (2013). Socioeconomic status and the health of youth: A
multilevel, multidomain approach to conceptualizing pathways. Psychological
Bulletin, 139, 606. https://doi.org/10.1037/a0029416
Sex differences in the association between marital status and the risk of cardiovascular,
cancer, and all-cause mortality: A systematic review and meta-analysis of 7,881,040
individuals. Global Health Research and Policy, 5, (2020), 4. https://doi.org/10.1186/
s41256-020-00133-8
Shelton, N. J., & Herrick, K. G. (2009). Comparison of scoring methods and thresholds of
the general health questionnaire-12 with the Edinburgh postnatal depression scale in
English women. Public Health, 123, 789–793. https://doi.org/10.1016/j.
puhe.2009.09.012
Solar, O., & Irwin, A. (2010). A conceptual framework for action on the social determinants
of health (policy and practice No. 2). WHO. https://www.who.int/publications-detail-
redirect/9789241500852.
Statista. (2023). Average age at marriage in England and Wales 1851-2020, by gender.
Retrieved from https://www.statista.com/statistics/557962/average-age-at-marria
ge-england-and-wales/. (Accessed 1 July 2023).
Tatangelo, G., McCabe, M., Campbell, S., & Szoeke, C. (2017). Gender, marital status and
longevity. Maturitas, 100, 64–69. https://doi.org/10.1016/j.maturitas.2017.03.002
The socioeconomic gradient in epigenetic ageing clocks: Evidence from the multi-ethnic
study of atherosclerosis and the health and retirement study. Epigenetics, 17, (2022),
589–611. https://doi.org/10.1080/15592294.2021.1939479
Thomas, A., Belsky, D. W., & Gu, Y. (2023). Healthy lifestyle behaviors and biological
aging in the US national health and nutrition examination surveys 1999-2018. The
Journals of Gerontology: Series A. https://doi.org/10.1093/gerona/glad082. glad082.
Umberson, D., & Thomeer, M. B. (2020). Family matters: Research on family ties and
health, 2010 to 2020. Journal of Marriage and Family, 82, 404–419. https://doi.org/
10.1111/jomf.12640
Understanding society. Waves 2 and 3 Nurse Health Assessment, 2010(2012), (2022), 38.
https://www.understandingsociety.ac.uk/sites/default/files/downloads/docu
mentation/health/user-guides/7251_User_Guide_Health_Assmt_w2_w3.pdf.
Waite, L., & Gallagher, M. (2001). The case for marriage: Why married people are happier,
healthier and better off financially. Crown.
Whisman, M. A., Robustelli, B. L., & Sbarra, D. A. (2016). Marital disruption is associated
with shorter salivary telomere length in a probability sample of older adults. Social
Science & Medicine, 157, 60–67. https://doi.org/10.1016/j.socscimed.2016.03.029
Williams, K., & Umberson, D. (2004). Marital status, marital transitions, and health: A
gendered life course perspective. Journal of Health and Social Behavior, 45(1), 81–98.
https://doi.org/10.1177/002214650404500106
Yu, Y.-L. (2023). Current marital status and epigenetic clocks among older adults in the
United States: Evidence from the health and retirement study. Journal of Aging and
Health, 35(1–2), 71–82. https://doi.org/10.1177/08982643221104928
Yu, Y.-L., & Liu, H. (2021). Marital quality and salivary telomere length among older men
and women in the United States. Journal of Aging and Health, 33(5–6), 300–309.
https://doi.org/10.1177/0898264320980250
Zarulli, V., Barthold Jones, J. A., Oksuzyan, A., Lindahl-Jacobsen, R., Christensen, K., &
Vaupel, J. W. (2018). Women live longer than men even during severe famines and
epidemics. Proceedings of the National Academy of Sciences, 115, E832–E840. https://
doi.org/10.1073/pnas.1701535115
Zhang, Y., & Liu, C. (2022). Evaluating the challenges and reproducibility of studies
investigating DNA methylation signatures of psychological stress. Epigenomics, 14(7),
405–421. https://doi.org/10.2217/epi-2021-0190
Zhang, Z., Liu, H., & Seung-Won, E. C. (2021). Marital loss and risk of dementia: Do race
and gender matter? Social Science & Medicine, 275, Article 113808. https://doi.org/
10.1016/j.socscimed.2021.113808